Multistate stochastic network interdiction via reliability modelling and evolutionary optimization

This paper extends the current stochastic network interdiction problem (SNIP) approaches to allow for the inclusion of multistate behaviour of link flow when a given link is targeted for interdiction. Under this modelling scenario, described as the multistate SNIP, a new optimization model is described for maximizing the reliability associated with an interdiction strategy constrained to a prespecified s-t flow and a prespecified cost. An evolutionary algorithm, known as the probabilistic solution discovery algorithm, is used to obtain quasi-optimal solutions for the multistate SNIP problem. This algorithm comprises a three-step process that implements multistate network reliability computation techniques along with an evolutionary optimization routine to find interdiction strategies. The solutions to different test cases show that the optimization routine is able to identify high-quality solutions in a significantly reduced search space.